therory of detonation waves - geocities therory of detonation waves 1 by j. von neumann summary -...
TRANSCRIPT
1
THERORY OF DETONATION WAVES1
By
J. VON NEUMANN
Summary - The mechanism by which a stationary detonation wave maintains
itself and progresses through the explosive is investigated.
Reasons are found which support the following hypothesis: The detonation
wave initiates the detonation in the neighboring layer of the intact explosive by
the discontinuity of material velocity which it produces. This acts like a very
vehement mechanical blow (~ 1,500 m/sec), and is probably mom effective than
high temperature.
The velocity of the detonation wave is determined by investigating all
phases of the reaction. and not only (as usually done heretofore) the completed
reaction. The result shows when the so called Chapman-Jouguet hypothesis is
true, and what formulae are to be used when it is not true.
Detailed computations win follow.
Objectives, Methods and General Principles of this Report
1. The purpose of this report is to give a consequent theory of the mechanism by
which a stationary detonation wave maintains itself and progresses through an
explosive. Such a theory must explain how the head of the detonation wave initiates
the reaction (and the detonation) in the intact explosive, and how a well-determined
constant velocity of this wave arises. Both aims can be achieved only after
overcoming certain characteristic difficulties.
2. Regarding the first objective, this ought to be said: We do not undertake to give a
theory of the initiation of a detonation in general. The viewpoints which we bring up
may throw some light on that question too, but our primary aim is to understand the
mechanism of the existing stationary detonation wave. Here the detonation of each
layer of the intact explosive is initiated by the stationary detonation wave which has
already engulfed its neighboring layer.
In gaseous explosions this may be explained by the high temperature of the
extremely compressed gas in the detonation zone. In solid explosives this
explanation is hardly available: The detonation wave is very narrow in space and
moving with very high velocity, so the chemical reaction (which supports the
detonation) cannot be treated as instantaneous. Hence the head of the detonation
wave corresponds to the just incipient reaction, and contains accordingly only an
1 A progress report to April, 1942 Division B National Defense Research Committee of the Office of
Scientific Research and Development Section B-1 OSRD 549 (May 4, 1942).
2
infinitesimal fraction of gas. The not-yet-reacted solid explosive in it is presumably
cold. At any rate the high gas temperatures in the wave head no longer dominate the
picture, and therefore cannot detonate the intact explosive.
We shall see that there is more likelihood in this: There must be a discontinuous
change of velocity of the bulk of the substance which crosses the wave head. This
acts as a violent blow, delivered at velocities of ~1,500 m/sec under typical
conditions - which is probably at least as effective as temperatures of
~2,500°- 3,000 ℃.2
3. As to the second objective, the present literature is dominated by the so,
called Chapman-Jouguet hypothesis. This hypothesis provides a definite
value for the velocity of the detonation wave, but its theoretical foundations
are not satisfactory. The experimental evidence is altogether reasonably
favorable to this hypothesis, but it is not easy to appraise because of the
slight information we possess concerning the physical properties of the
substances involved under the extreme conditions in a detonation.
All theories on this subject are based on the Rankine-Hugoniot
equations of a shock wave, which are actually applications of the
conservation theorems of mass, momentum, and energy. The
Chapman-Jouguet hypothesis is based on a consideration of these principles
in the completed reaction only.
We shall apply them to all intermediate phases of the This necessitates
the investigation of the entire family of all so called Rankine-Hoguniot
curves, corresponding to all phases of the reaction. By doing this we shall
succeed in determining the velocity of the detonation wave. We find that the
Chapman-Jouguet hypothesis is true for some forms of the above-mentioned
family of curves, and not for others. We obtain precise criteria, which
determine when it is true, and also a general method to compute the velocity
of the detonation wave which is always valid.
4. All these discussions are made in a general, as far as feasible qualitative,
way, avoiding detailed computations. Specific computations which determine
the detonation velocity for definite types of explosives will be made
subsequently. In this connection the question is of importance whether the
configuration of the family of curves mentioned above, for which the
Chapman-Jouguet hypothesis fails to be true, ever occurs for real explosives.
This question will be considered together with the first-mentioned
computations. In all these discussions the compressibilities and specific heats
2 We are thus disregarding the possibility that the detonation is propagated by special kinds of
particles (ions, etc.), moving ahead of the wave head. Indeed, if the views which we propose are
3
of solids and gases under detonation conditions are the decisive factors.
The properties of a detonation which has not yet reached its stationary
wave stage will also be investigated later. It is to be hoped that this will
connect the present theory with the difficult questions of initiating a
detonation: of “primers” and “boosters”.
The present work is restricted to plane waves in absolutely confined
explosives. The effects of spatial expansion on spherical waves and in the
case of partial confinement will be considered later.
The author finally wishes to express his thanks to E. Bright Wilson, Jr.,
and to R. H. Kent, for whose valuable suggestions and discussions he is
greatly indebted.
§§1-3. Formulation of the Problem
1. In studying the mechanism of a detonation the following approximate
picture suggests itself.
Imagine that the detonation wave moves across the explosive in parallel
planes, i.e., everywhere in the same direction. Choose this direction as the
negative x-axis. Then conditions are constant in all planes parallel to the y,
z-plane. The explosive must be confined by some cylindrical boundary, i.e.,
by the same boundary curve in each one of the above parallel planes. Assume
this confinement to be absolutely unyielding. Assume that all motions
connected, with the detonation take place in the direction of the x-axis, i.e.,
that there are no transversal components parallel to the y, z-plane.
The limitations of this picture are obvious, but it is useful for a first
orientation. It ought to be essentially correct for a heavily confined stick of
explosive, detonated at one end surface.3
Under these assumptions the process of detonation can be treated
one-dimensionally; i.e., we may disregard the coordinates y, z and treat
everything in terms of the coordinate x and of the time t.
2. We restrict ourselves further by assuming also that the detonation wave
has reached its stationary stage; that is, that it moves along without any
change of its structural details.
Then, among other things, the wave velocity must be constant in time.
found to be correct, no such special particles will be needed to explain the detonation wave. 3 It may even be approximately applicable to an unconfined stick of explosive, since the great
velocity of the detonation wave - i.e., the brevity of the available time interval - makes the inertia of a solid explosive itself act as a confinement.
In a higher approximation, however, lack of confinement causes corrections which must be determined. This problem was considered by G. I. Taylor and H. Jones in the British reports RC
193 and RC 247 (1941). The authors, however, used a picture of the mechanism of detonation
that differs from the one we shall evolve. It is proposed to take up this subject from our point of
4
We now modify the frame of reference by making it - i.e., the origin of the
coordinate - move along with the wave. To be specific, let the origin of the
x-coordinate be at every moment at the head of the wave. The intact
explosive is to the left of this. In other words:
(2-A) The intact explosive occupies the space x < 0.
(2-B) The detonation - i.e., the chemical reaction underlying it-sets in at x = 0.
This is the head of the detonation wave.
(2-C) The entire process of detonation - i.e., the chemical reaction mentioned
above - occurs in successive stages in the space x > 0. The part of this space
which it occupies is the reaction zone. The remoter parts are occupied by the
completely reacted (burnt) products of the detonation,4 the burnt gases. This
is the region behind the detonation wave.
3. The detonation wave has a constant velocity with respect to (and directed
toward) the intact explosive, say D. Since the detonation wave is at rest in
our frame of reference, this means that the intact explosive has the velocity D
with respect to (and directed toward) the fixed head of the wave. That is, at
every point x < 0 (in (2-A)) the velocity of matter is D. (D > 0, in the
direction of the positive x-axis!)
At every point x > 0 (in (2-C)) we have a velocity of matter u = u(x) and a
fraction n = n(x) > 0, ≤ 1 expressing to what extent the chemical reaction has
been completed there. (That is at x > 0 a unit mass contains n parts of burnt
gas and 1 - n parts of intact explosive.) During the time dt matter in this
region moves by dx = udt. So if the reaction velocity (under the physical
conditions prevailing at x) is )(xαα = then we have5
u
n
dx
dn α)1( −= . (3.1)
As x increases from 0 toward +∞ , n increases from 0 toward 1.
According to the details of the situation, n may or may not reach the value
(completely burnt gas) for a finite x (cf. previous footnote). We assume for
the sake of simplicity that the former is the case, i.e., that the reaction is
exactly completed for a finite x.
To conclude the description of condition in our phenomenon: At every
point x < 0 (in (2-A)) we have the same physical characteristics - say
pressure 0p and specific volume 0v . These describe the intact explosive.
view in a subsequent report. 4 This may be true exactly, or only asymptotically. 5 In assuming the existence of a well defined and stationary reaction velocity (i.e., of one which
is a numerical constant in time), a definite physical hypothesis is being made: That of (at least) kinetic quasi equilibrium at every point of the reaction. This hypothesis is, however, a natural
one to make, unless there is definite evidence to the contrary; And there does not seem to be any
so far.
5
At every point x > 0 (in (2-C)) we have a pressure p = p(x) and a specific
volume v = v(x).
The nature of the chemical reaction is expressed by a functional
relationship
),,( vpnA=α , (3.2)
where we assume A(n, p, v) to be a known function. The stability of the intact
explosive requires
0),,0( 00 =vpA . (3-3)
The nature of the explosive and its mixtures with the burnt gas is
expressed - as far as it interests us - by its caloric equation for each 0≥n ,
1≤ , i.e., by the functional relationship determining its inner energy, per unit
mass
),,( vpnEe = , (3.4)
where we assume E(n, p, v) to be a known function.
§§4-7. Discussion of the Equations. The Chapman-Jouguet Hypothesis
4. The substance which passes from the intact explosive state-the space x < 0
in (2-A) - through the various phases of the chemical reaction underlying the
detonation - the space x > 0 in (2-C) - must fulfill the conditions of
conservation of mass, momentum, and energy. Indeed these are the
mechanical conditions for the stationarity of the detonation wave.6 Denote
the mass flow - the amount of matter crossing the wave head per second,7
i.e., the amount of matter detonating per second, by µ . Then the conditions
of conservation become:
Mass: 0vD µ= , vu µ= , (4.1)
Momentum: 0)( ppuD −=−µ , (4.2)
Energy: 02
002
),,(2
1),,0(
2
1DpupvpnEuvpED −=
−−+µ .
(4.3)
(4.1), (4.2), (4.3) together with (3.1), (3.2), determine everything.
But (3.1), (3.2) merely establish the scale of conversion of the two
variables x and n into each other. If we are satisfied to use n - instead of x - as
the independent variable, then we can rely upon (4.1), (4.2), (4.3) alone. If
we obtain from them p = p(n), v = v(n) and u = u(n), then (3.1), (3.2) may be
stated as
6 This is the classical procedure of Rankine and Hugoniot. The application to all intermediate
phases of the reaction for the purpose of a structural analysis was suggested by G.. I. Taylor and H. Jones loc. cit. (footnote 3). 7 And per unit surface of the y, z-plane.
6
⌡⌠
−=
n
vpnAn
udnx
0 ),,()1(, (4.4)
i.e., as a conversion formula in the above sense.
This is the procedure which we shall use.
5. The algorithm of solving (4.1), (4.2), (4.3) is well known.
We consider 0p , 0v as given, D, µ as unknown parameters of the
problem,8 and p, v, u as unknown functions of n.
9 (4.1) expresses D, u in
terms of µ , v ( 0v is given!), so that equations (4.2), (4.3) remain. These
are easily transformed into
2
0
0 µ=−−vv
pp, (5.1)
),,(),,0())((2
10000 vpnEvpEvvpp ==−+ . (5.2)
Thus µ determines, for every n > 0, 1≤ , p, v by the implicit equations
(5.1), (5.2).
The strange thing in all this is that one unknown parameter - say µ -
remains undetermined. But the stationary state of the detonation wave of a
definite chemical reaction ought to possess - if it exists at all -
unambiguously determined characteristics. Hence it should be possible to
formulate some further condition which completes the determination of µ .
Before we consider this question, however, let us return to (5.1), (5.2).
Their solution can be illustrated by a familiar graphical method (Fig. 1).
Plot the curve (5.2), the Rankine-Hugoniot curve, in the p, v-plane,
together with the point ( 0p , 0v ). Denote the angle between the direction of
the negative v-axis and the direction ( 0p , 0v )→ (p, v) by φ . Then (5.2)
states that p, v lies on the Rankine-Hugoniot curve, and (5.1) states that
φµ tan= , φtan0vD = . (5.3)
Thus the unknown parameter φ is substituted for the unknown
parameter µ (or D).
This picture should be drawn for each n > 0, 1≤ , with the same ( 0p ,
0v ) and φ but varying Rankine-Hugoniot curves, and so the corresponding
p, v are obtained.10
Observe that all these considerations are also valid for n = 0, in which
8 Not dependent upon x or n! 9 Instead of x, cf. the end of §4. 10 And the φtanvu = .
7
case they describe the conditions under which a discontinuity in our
substance can exist - without making use of any chemical reaction at all. This
phenomenon is particularly important in gases and in liquids. It is known as
shock wave, and is actually an essential component of the theory of the
detonation wave.
6. While φ is undetermined, it is not entirely arbitrary (Fig. 2).
To begin with, (5.3) necessitates 0tan >φ , so φ must lie in the
Quadrant I or III. We have accordingly:
Quadrant I: 0pp > , 0vv < , hence u < D.
Quadrant III: 0pp < , 0vv > , hence u > D.
In Quadrant I the burnt gases are carried along with the detonation wave
(i.e., D - u > 0); their density and pressure are higher than those in the
explosive. This is detonation proper.
In Quadrant III the burnt gases are streaming out of the explosive (i.e., D
- u < 0); their density and pressure are lower than those in the explosive. This
is the process commonly known as burning.
8
We are interested in the process of detonation only, so we assume φ in
Quadrant I.
Now our treatment of the chemical reaction (cf. in particular footnote, p.
206) compels us to postulate for each n > 0, 1≤ the existence of a well
defined physical state, fulfilling the requirements of the theory of §§4-5.
Therefore φ must be at least the angle nφ of the tangent from ( 00 , vp ) to
the Rankine-Hugoniot curve. For nφφ < the line on which (p, v) should lie
does not intersect that curve at all. For nφφ ≥ the situation is usually this: If
nφφ = then there exists exactly one value for (p, v) - the tangent point X; if
nφφ ≥ then there exist exactly two values for (p, v) - the lower intersection
point Y and the upper intersection point Z. We assume these qualitative
conditions - as exhibited by Fig. 2 - to hold for the Rankine-Hugoniot curves
of all n.
7. These considerations were originally applied to n = 1 only. (Cf. however
footnote, p. 206.) The lower limit for φ is then 1φφ = - the direction of the
tangent to the Rankine-Hugoniot curve n = 1.
The classical theory of detonation is based on the assumption that the
actual value of φ : is this lower limit 1φ . This is the so-called hypothesis of
Chapman and Jouguet.11
Various theoretical motivations have been
proposed for this hypothesis, mostly based on considerations of stability.12
The experimental evidence is not easy to appraise, since the question of the
validity of this hypothesis is intertwined with uncertainties concerning the
caloric equations of the substances involved under the extreme conditions in
a detonation.
We propose to carry out a theoretical analysis of the Chapman-Jouguet
hypothesis by a study of the Rankine-Hugoniot curves for all n 0≥ , 1≤ , i.e.,
for all intermediate phases of the reaction. It will be seen that a proper
understanding of the situation with the help of the curve n = l alone - as
attempted always heretofore - is impossible. It is necessary to consider the
family of all curves n 0≥ , 1≤ . By doing this we shall supply the missing
condition mentioned in §5 (after (5.1), (5.2)), i.e., determine φ (that is µ ,
D).
For certain forms of the family of all curves n 0≥ , 1≤ the
Chapman-Jouguet hypothesis will turn out to be true; for other forms it is
11 For this, and several references in what follows, cf. the report of G. B. Kistiakowsky and E.
Bright Wilson, Jr., “The Hydrodynamic Theory of Detonation and Shock Waves”, OSRD (I941).
This report will be quoted as K.W. Concerning the Chapman-Jouguet hypothesis, cf. K.W., Sec. 5.
9
false and another (higher) value of φ : will be found.
§§8-9. General Remarks
8. The discussion at the end of 6 showed that we must have nφφ ≥ for all
n 0≥ , 1≤ . Now the Chapman-Jouguet hypothesis is based on the
consideration of the Rankine-Hugoniot curve n = 1 alone, and it consists of
assuming 1φφ = . Hence it is certainly unacceptable unless nφφ ≥1 for all
n 0≥ , 1≤ . That is: Unless nφ assumes its maximum (in 10 ≤≤ n ) for n =
1. Or in a more geometrical form: Unless every line issuing from the point
( 00 , vp ) which intersects the curve n = 1 also intersects all other curves
n 0≥ , 1≤ .
We illustrate this condition by exhibiting two possible forms of the family
of all curves n 0≥ , 1≤ (Figs. 3, 4).
The condition is fulfilled in Fig. 3, but not in Fig. 4.
12 For a summary cf. K.W., Sec. 5, pp. 8-11.
10
From a purely geometrical point of view these two forms are not the only
possible ones for the family of curves n 0≥ , 1≤ . We shall not attempt to give
here a complete enumeration. The main question in this connection is,
however, which forms of this family occur for real explosives. We shall take
up this question in a subsequent investigation.
9. We shall obtain the missing condition for φ repeatedly mentioned before.
This derivation will differ essentially from the existing analyses, since those
make use of the curve n = 1 only. (For this, and for some of the remarks
which follow, cf. footnote, p. 209.) There is nevertheless one element in these
discussions which deserves our closer attention.
The discussions referred to treat the upper intersection points Z and the
lower intersection points Y separately.13 That is, in order to establish the
Chapman-Jouguet hypothesis - i.e., the tangent point X - the Z and the Y are
ruled out by two separate arguments. We saw in §8 that the
Chapman-Jouguet hypothesis cannot be always true; hence the arguments in
question cannot be conclusive.14
Our considerations will prove that the
exclusion of the Z on the curve n = 1 must be maintained, but not that of the
Y.
Thus the first part of our analysis will be restricted to the curve n = l and
on it to the upper intersection points Z: proving that they cannot occur. In
doing this we shall use a line of argument which is closely related to the
traditional one referred to above - even to the detail of being based on a
comparison of the detonation velocity to the sound velocity in the burnt gas
behind it.15 In spite of this similarity, however, the two argumentations are
not the same: The traditional one is, as mentioned before, essentially one of
stability, while ours is purely cinematical. It is actually closer to a viewpoint
which has been put forward recently by G. I. Taylor.16
In other parts of our analysis we shall have to consider all curves
n 0≥ , 1≤ and their positions relative to each other. This is essentially
different from the existing analyses referred to above.
On one occasion we shall consider the possibility of a shock wave in the
reaction zone and the change of entropy which it causes. This phenomenon has
been investigated in connection with the Chapman-Jouguet hypothesis.17 but in a
different arrangement: In a stability consideration, in order to exclude (on the
13 Cf. our Fig. 2, or K.W., Fig. 5-1, p. 8. 14 Cf. also K.W., bottom of p. 11. 15 Cf. K. W., pp. 8-9. 16 British report of G. I. Taylor, “Detonation Waves", RC l78 (1941). This report will be quoted
as T. 17 Cf. K.W., pp. 9, 11.
11
curve n = 1) the lower intersection points Y also, to prove the Chapman-Jouguet
hypothesis. Our procedure again is not one based on stability, and besides we use
it for a different purpose. Furthermore we shall find that under certain conditions
the lower intersection points cannot be disregarded.
§§10-12. Conditions in the Completely Burnt Gas
10. Let us consider the state (p, v) at n = 1, which is a point on the
Rankine-Hugoniot curve n = 1. This is the back end of the reaction zone, where
the chemical reaction is just completed (cf. (2-C)). We use the frame of reference
in which the detonation wave – and hence this particular point too - is at rest.
Consider now the conditions behind this point P (Fig. 5):
In order to complete the picture we must remember that the explosive was
assumed to be completely confined (cf. 1); hence it is logical to assume a back
wall behind it, say at Q.18 This back wall is at rest in the original frame of
reference in which the intact explosive is at rest; in our present frame of reference
the intact explosive has the velocity D (cf. the beginning of 3), and therefore the
same is true of the back wall Q.
So the burnt gases lie between the points P, Q, which move with the
velocities 0, D, respectively. The gases stream across P with the velocity
φtan=u (cf. footnote, p. 208), while Q is an enclosing wall.
The phenomenon is certainly not stationary in P, Q, since P and Q recede
from each other; but we assumed it to be stationary in the reaction zone O, P. In
P, Q, on the other hand, no chemical reaction is going on : The gases there are
completely burnt.
The question is whether the necessity of fitting a gas-dynamically possible
state of motion between P and Q imposes any restriction on the state at P, i.e., on
the intersection point on the curve n = 1.
An answer can be obtained mathematically by integrating the differential
equations of gas dynamics in P, Q.19 But it can also be found by more qualitative
considerations, which we shall now present.
18 If the explosive were unconfined at Q, the conclusions of these sections would probably be valid a
fortiori, since they demonstrate the necessity of a rarefaction wave in the burnt gases under certain conditions (cf. §12). But we prefer to restrict ourselves to the case of absolute confinement. 19 This is actually contained in the computations of T, pp. 1-4.
12
11. Consider the stretch O, Q in the original frame of reference, in which the
intact explosive and the back wall at Q are at rest. Throughout the entire period
of time in which the detonation progressed from its start at Q to its present head
at O the explosive to the left of O remained intact. Hence no substance was
transferred from one side of O to the other,20 and so the total mass in O, Q was
not changed. Therefore the average density in O, Q was not changed either - i.e.,
it is now, when the wave head is at O, the same as it was when the detonation
started at Q - i.e., the same as in the intact explosive.
The specific volumes in O, P are the v on the line of Fig. 2, from ( 00 , vp ) to
X or Y or Z, as the case may be; hence all smaller than the specific volume 0v
of the intact explosive. That is the density in O, P is everywhere greater, as in
the intact explosive.
Hence the average density in P, Q is lower, as in the intact explosive, and a
fortiori as at P. Consequently we can assert:
The density at P is greater than in some places in P, Q.
Let P' be the place (in P, Q) where the density begins to fall below its value
at P.21 The density, i.e., the specific volume v, is therefore constant in P, P'.
12. We now return to the frame of reference of §§3-5 and 10, where the wave
front O - and so the geometrical locus of the reaction zone O, P-is at rest.
v is constant in P, P' hence the same is true of u22 and p,
23 and, along with p,
v, of the sound velocity c.24
By its definition, P' is the head of a rarefaction wave looked at from O, P, P'.
Hence moves with sound velocity towards O, P, P'; i.e., the velocity of P' is u -
c. We know that we may take this u - c at P instead of P', and that u - c taken at
P is constant in time.25 Hence u - c < 0 would imply that P' will reach and enter
the zone O, P in a finite time, thus disturbing its stationarity. Hence u - c 0≥ ,
i.e., c ≤ u at P.
Let us now consider Fig. 1, taking n = 1 for its curve. φtan=u (cf.
footnote, p. 208), and c too can be expressed in terms of this figure. Denote the
angle between the direction of the negative v-axis and the direction of the
tangent of the curve at (p, v) by ψ . We claim that ψtan=c . This can be
20 O is now at rest with respect to the intact explosive! 21 P’ is P, or to the right of P. 22 Because of the conservation of mass, i.e., (4.1) or footnote, p. 208. 23 Because of the adiabatic law, which holds throughout P, Q, since there are no chemical reactions
or shock waves there. 24 We measure c with respect to the gas, which itself moves with the velocity u. 25 The zone O, P is stationary!
13
established by a direct thermodynamical computation,26 and also by a more
qualitative argument which we shall give now.
The line ( 00 , vp )→ (p, v) represents a detonation wave ending with the
burnt gas. Therefore a line ( ',' vp )→ (p, v) where both points (p', v') and (p, v)
lie on the curve n = 1, represents a discontinuity which begins and ends in the
burnt gas, i.e., a shock wave in it.27 If (p', v') moves very close to (p, v), then
this direction tends to the tangent at (p, v). And the shock wave tends to a very
small discontinuity, i.e., it becomes very weak. Now the velocity of a very weak
disturbance is very nearly sound velocity. Hence c, ψ are connected in the
burnt gas in the same way as D, φ were in the intact explosive. So the
φtan0vD = of (4.1) becomes ψtanvc = .
Thus our uc ≤ becomes φψ tantan vv ≤ , i.e., φψ ≤ : In other
words:
The direction ( 00 , vp )→ (p, v) cannot be less steep than the direction of
the tangent to the Rankine-Hugoniot curve n = l - i.e., the direction of the
curve itself - at (p, v).
One look at Fig. 2 suffices to show that this means the exclusion of the
upper intersection points Z on the curve n = 1.28 We restate this:
The state p(n), v(n) at n = 1, which lies on the curve n = 1, cannot be an
upper intersection point.
§§13-15. Discontinuities and Continuous Changes in the Reaction Zone.
Mechanism which Starts the Reaction.
13. We turn next to the consideration of the Rankine-Hugoniot curves for all n
0≥ , 1≤ .
The state at the point of the reaction zone with a given n is characterized by
the data p(n), v(n) and u(n) (cf. the end of §4). These points p(n), v(n) for n 0≥ ,
1≤ form a line Λ in the p, v-plane, the line representing the successive states
across the reaction zone (cf. the end of §l2).
This line Λ must intersect the curve of every n 0≥ , 1≤ . If this
intersection occurs in a (unique) tangent point, then that is p(n), v(n); if it occurs
in two intersection points - the upper and the lower - then one of those is p(n),
v(n). Let us now follow the history of p(n), v(n) as n increases from 0 to 1.
To begin with, there is no absolute reason why p(n), v(n) should vary
26 Cf. K. W. pp. 9-11. 27 Involving no chemical reaction! 28 As mentioned in footnote 19, the computations of T, pp. 1-4, prove the same thing. The state in P,
Q (cf. Fig. 5) for the tangent point X or a lower intersection point Y(cf. Fig. 2) is exhibited by Figs. 1b
and 1a, respectively, in T.
14
continuously with n. Consider now an 0n where they are discontinuous.
Follow the path of matter (in the detonation) which crosses that point of the
reaction zone. It moves in the direction of the progressing chemical reaction, i.e.,
of an increasing n. The discontinuity of p(n), v(n) necessitates that these
quantities have different limiting values, as 0n is approached on the incoming
side from below, and on the outgoing side from above; Both limiting positions
of p(n), v(n) must be intersections of Λ with the curve 0n . So they are the
upper and the lower intersection point;29 the question is only: which is which?
The two intersection points of Λ on the curve 0n represent the two sides
of a shock wave;30 indeed our above description of the situation at n; makes it
clear that there is a shock wave at the point no of the reaction zone.
Now it is well known - for thermodynamical reasons, but also intuitively
plausible - that in a shock wave the high-pressure area always absorbs the
substance of the low-pressure area, i.e., that matter passes from the low-pressure
state into the high-pressure state.31
In the present case matter moves in the direction of increasing n (cf. above).
Therefore the lower intersection point is the limiting position when 0n is
approached from below, and the upper intersection point is the limiting position
when 0n is approached from above.
Summing up:
The point p(n), v(n) varies continuously, except for possible jumps from
the lower to the upper intersection point which occur - if at all - always in this
direction with increasing n.
14. Let us consider the conditions at n = 0, i.e., at the head of the wave. The
state there adjoins the ( 00 , vp ) of the intact explosive which lies on the curve n
= 0.
Assume first that the variation of the p(n), v(n) at this point is continuous,
i.e., that ( 00 , vp ) is the limiting position when n approaches 0 from above.
In this case there is no agent to start the chemical reaction which supports
the detonation. This reaction ought to set in, and quite vehemently, at n = 0, i.e.,
for small values of n > 0. Or, if we use the independent variable x: at x = 0, i.e.,
for small values of x > 0. Now the assumed continuity means that the conditions
in this critical zone - the beginning of the reaction zone - differ only
insignificantly from those at ( 00 , vp ), i.e., throughout the intact explosive (in x
29 And we do not have the tangent position! 30 Cf. a similar discussion in §l2. 31 The thermodynamical reason is that the entropy is higher on the high-pressure side. Cf. K.W., p.
11. For a discussion of this property of shock waves, cf. e.g. J. W. Rayleigh, “Aerial Plane Waves of
15
< 0). Thus the reaction cannot start in this region. If it did - for any reason
whatever - it should a fortiori have done so in the intact explosive (in x < 0).
There was much more time available there, and yet, by assumption, no reaction!
This argument is qualitative, to be sure, but it is quite easy to amplify it
mathematically.32
Thus we must have a discontinuity at n = 0. By the result of §l3 this means:
( 00 , vp ) must be the lower intersection point, and as n increases (from 0
on), p(n), v(n) immediately jumps to the upper intersection point - say
( 00 ,vp ) - and goes on continuously from there.
Thus the reaction zone sets in with a shock wave, an abrupt increase of p,
and with it an equally abrupt decrease of v (cf. (5.1)) and of u (cf. footnote 10).
These abrupt changes of (p, v) may start the reaction, particularly because
they imply usually an increase of the temperature. But the change of u is even
more remarkable. This decrease from D to 0u means, of course, that the intact
explosive (in x < 0) receives a vehement blow, delivered by the wave head with
the velocity.
0uDw −= .
Using (5.1), (5.3) and footnote 10, we see that
))(()(10
0000
00
00
vvppvvDv
vuDw −−=−=
−=−= µ .
(14.1)
This velocity is smaller than, but in the order of magnitude of, the detonation
velocity D. It is comparable to the thermic agitation of a very high temperature;
indeed it may be more effective, since it is delivered with one systematic
velocity, and not in statistical disorder.33 Consequently the discontinuity of the
velocity provides at any rate the mechanism needed to start the reaction.34
15. We know from §13 that p(n), v(n) is immediately after a discontinuity at an
upper intersection point. We also know from §l2 that the latter cannot be the
case at n = 1. We have:
p(n), v(n) is continuous at n = 1, in the position indicated in §l2.
Finite Amplitude". Scientific Papers, Vol. 5, Cambridge 1912, particularly pp. 590-1. 32 Consider the formula (4.4) which expresses x in terms of n. If we have continuity, i.e., if n→ 0
implies 0pp→ ,
0vv→ , then (3.3) yields A(n, p, v) → 0. So we must expect divergence of the
integral (4.4) (near n = 0), and failure to obtain an acceptable x.
This is, of course, not essentially different from the text's verbal argument: We cannot make x→ 0 -
nor even x → finite for n→ 0, i.e., the reaction cannot be started on any finite interval. 33 For a typical high explosive, like TNT, D ~ 6,000 m/sec, w ~ 1,500 m/sec, so that the
corresponding temperature would be ~ 2,500°-3,000° C. 34 This view would necessitate a modification of (4.4), but in a favorable sense: towards even
smaller changes of x in the neighborhood of n = 0.
16
We are now informed about the behavior of p(n), v(n) at n = 0 and n = 1, and
also at discontinuities for n >0, <1. There remains only the necessity of
discussing its behavior when it varies continuously near an n >0, <1. We are
particularly interested how - if at al1 - it can change under these conditions from
an upper intersection point to a lower one, or vice versa.35
Let us therefore consider such a change. If p(n), v(n) changes at 0nn =
( 0n >0, <1) continuously from an upper intersection point to a lower one, or
vice versa, then it is necessarily a tangent point at 0n .
At this juncture it becomes necessary to introduce the envelope of the family
of all curves ≥n , 1≤ . This envelope may or may not exist (cf. Figs. 4 and 3
for these two alternatives, respectively). At any rate, if the tangent point which
we are now considering is not on the envelope,36 then the position of the curves
for the n near to 0n is this (Fig. 6).
Thus the curves with 0nn > are all on one side of the curve and the
curves with 0nn < are all on the other side. Consequently the line Λ does
not intersect that one of these two curve systems which is on the concave
side of the curve no. But Λ must intersect the curves of all n (cf. §6). So we
have a contradiction.
Hence the envelope must exist, and our tangent point must lie on it. At
this point Λ is tangent to the curve 0n . But the envelope has at each one
of its points the same tangent as that curve n of the family which touches it
there. Hence Λ is also a tangent for the envelope.
35 The first change (for n increasing) actually excludes a discontinuity according to the result of §l3. 36 This is meant to include the case where the envelope does not exist.
17
Summing up:
If p(n), v(n) changes at 0nn = ( 1,00 <>n ) continuously from an
upper intersection point to a lower one (cf. footnote 35), then:
(A) The envelope of the family of all curves n exists.
(B) The point p( 0n ), v( 0n ) lies on the envelope.
(C) The line Λ is at this point tangent to both the curve 0n and to
the envelope.
It is easy to visualize that if the point p( 0n ), v( 0n ) fulfills (B), (C), then
the change from an upper to a lower intersection, or vice versa, can be
effected continuously (Fig. 7, Cases 1, 2).
But it is also possible to have (B), (C), and nevertheless no such change
(Fig. 7, Cases 3, 4).
§16. Conclusions
16. The results of §§l2, 13, 14 and the two results of §l5 permit us to form a
complete picture of the variations of p(n), v(n).
By §l4 p(n), v(n) begins, for small n > 0, as an upper intersection point. If
it stays one for all n > 0, < 1, then §12 and the first result of §l5 necessitate
that it terminate at n = 1 as a tangent point. In this case Λ is a tangent to the
curve n = 1.
If the above assumption is not true, then p(n), v(n) must change (as n
18
increases) from an upper intersection point to a lower one, for some n > 0, <
1. By §13 this must occur continuously (cf. also footnote 35), and by the
second result of §15 we have (A)-(C) there: The envelope exists, and Λ is
tangent to it at the point in question. In this case Λ is tangent to the
envelope.
So we see:
Λ is tangent either to the curve n = 1 or to the (then necessarily
existing) envelope.
Since the line Λ comes from the given point ( 00 , vp ) this condition
leaves only a finite number of alternatives for it, i.e., for its angle φ with
the direction of the negative v-axis. So we have obtained essentially the
missing condition, referred to at the beginning of §9 and before.
Indeed, when the envelope does not exist (as in Fig. 3), then we see that
Λ must be tangent to the curve n = l, proving the Chapman-Jouguet
hypothesis. We restate this:
If the envelope does not exist - i.e., if the curves for all n do not intersect
each other (cf. Fig. 3) - then the Chapman-Jouguet hypothesis is true.
On the other hand we saw a case in §8 in which the Chapman-Jouguet
hypothesis cannot be true. Then the envelope must exist, and Λ must be
tangent to it.
We leave it to the reader to discuss the details of the solution in various
typical cases, e.g., in Fig. 3 (the Chapman-Jouguet hypothesis is true), and in
Fig. 4 (the Chapman-Jouguet hypothesis is not true). The results referred to at
the beginning of §16 are all that is needed in all these cases.